Build a Plex box

A Plex box has one hard problem and a few easy ones. The hard problem is transcoding - reshaping a video on the fly so it plays on whatever device asked for it. Get that right and everything else is just disks and a network. Get it wrong and you buy a powerful, expensive machine that still stutters. Here is how to build one that just plays.

1. Know when you transcode (and when you do not)

If every device in your house can play a file directly, Plex just copies the bytes across and almost any hardware works. You transcode when a device cannot handle the file - an old TV, a phone on cellular data, a lower bitrate for someone travelling. The more people watch remotely, the more transcoding you do, and the more the hardware matters.

So the real question is not "how fast is the CPU" but "how many streams might need reshaping at once." Answer that first.

2. The GPU is the whole game

Modern hardware transcoding is done by a fixed-function video engine, not the main CPU cores. This is the single most important fact in the build:

  • Intel QuickSync (built into most Intel iGPUs) is the sweet spot. It sips power, handles many simultaneous 1080p streams, and it is often already inside the mini PC you were going to buy anyway.
  • An NVIDIA GPU with NVENC handles heavier loads and 4K, at higher cost and power. Consumer cards cap the number of concurrent streams unless you lift the limit.
  • The CPU almost does not matter for transcoding. Do not pay for a monster CPU to play movies.

The practical upshot: a small box with a recent Intel iGPU handles most households for very little money and noise. The full comparison of which chips and cards transcode how many streams lives at ServerRigs - that is the page to open before you spend anything.

3. Pick the box

For most people the answer is a mini PC with a modern Intel iGPU. It is quiet, draws a handful of watts at idle, fits behind the TV, and QuickSync does the heavy lifting. Buy one with enough RAM for the OS and Plex, and you are done.

Step up to a discrete GPU only if you know you will push several 4K transcodes at once, or you are also using the machine for AI or gaming. Otherwise it is wattage and fan noise you will not use.

4. Storage: plan for the library, not today

Media libraries only grow. Two decisions:

  • Capacity. A 4K movie is tens of gigabytes; a serious library is measured in double-digit terabytes. Buy CMR NAS drives sized for where the library is heading, not where it is now. Which drives are CMR and which are quietly SMR is the thing NASdisks tracks; check before you buy.
  • Where it lives. Small library, a couple of drives inside the Plex box is fine. Larger, and you want a separate NAS holding the media, with the Plex box reading from it over the network. Keep the movies and the server loosely coupled so you can upgrade either one alone.

5. Do not starve the network

Direct play of a 4K remux is a firehose of data. On a wired gigabit link it is fine; over flaky Wi-Fi it is buffering. Put the Plex box and the storage on wired connections, and if the library sits on a separate NAS, a 2.5-gigabit link between them removes the last bottleneck. The switch-and-NIC side is covered at HomeSwitches.

6. Plex or Jellyfin?

Both play your library and both hardware-transcode. Plex is more polished and has better remote-access and app coverage, with some features behind a subscription. Jellyfin is fully free and open, with a rougher edge in places. If you want it to just work for the whole family, Plex. If you want no subscriptions and full control, Jellyfin. The hardware advice above is identical either way.

The short version

Transcoding is the only hard part, and a recent Intel iGPU with QuickSync solves it for most homes at tiny power and cost. Do not overspend on the CPU. Size CMR storage for the library's future, keep the media on a wired link, and pick Plex or Jellyfin to taste. Start with the transcode comparison at ServerRigs, then pick your drives at NASdisks.